Hi, Is it possible to "force" Gretl to show/compute simply a "0" instead of a "1.1102e-16"? For example, in the script below the d series holds, in theory, nothing else but 0-s but in practice it still holds some kinds of "1.1e-17" numbers nulldata 100 genr x=normal(0,1) genr f=1/(1+exp(-x)) genr g=exp(x)/(1+exp(x)) # f = g genr d=f-g # should be a series of 0 print d summary d Best, Artur _______________________________________________ Gretl-users mailing list -- gretl-users(a)gretlml.univpm.it To unsubscribe send an email to gretl-users-leave(a)gretlml.univpm.it Website: https://gretlml.univpm.it/postorius/lists/gretl-users.gretlml.univpm.it/

Am 28.10.2020 um 15:09 schrieb Ioannis Venetis: d = abs(d)<0.00001 ? 0 : d Hi, as a related function let me point out that in the "extra" addon (that is automatically shipped with gretl) we have zeroifclose() which is aimed at matrix arguments, however. There the default but settable threshold is 1e-12. Example: <hansl> include extra.gfn matrix m = {1e-17, 1e-8} zeroifclose(&m) print m </hansl> Στις Τετ, 28 Οκτ 2020 στις 3:45 μ.μ., ο/η Artur Bala < artur.bala.tn(a)gmail.com&gt; έγραψε: > Hi, > > Is it possible to "force" Gretl to show/compute simply a "0" instead of a > "1.1102e-16"? > For example, in the script below the d series holds, in theory, nothing > else but 0-s but in practice it still holds some kinds of "1.1e-17" > numbers > > But a much more fundamental point is that this is simply the old thing about floating-point arithmetic, where rounding and inexact representation occurs. In most cases for the computer there is no real difference between 1e-17 and 0. Therefore programming is best done a little differently from what the algebraic formulas tell us. There is a lot of classic literature about this. cheers sven _______________________________________________ Gretl-users mailing list -- gretl-users(a)gretlml.univpm.it To unsubscribe send an email to gretl-users-leave(a)gretlml.univpm.it Website: https://gretlml.univpm.it/postorius/lists/gretl-users.gretlml.univpm.it/

Am 28.10.2020 um 18:27 schrieb Artur Bala: Thank you Ioannis and Sven!!! To be frank, the kind of expression "abs(d)<0.00001 ? 0 : d" sounds to me not that much intuitive - anyway, less than zeroifclose() stuff - but probably it's just lack of experience. It's basically just a compact if-block which you can read out like this: Is abs(d) smaller than 1e-5 ? ( <- note the literal question mark in the syntax!) If yes, set it to zero - or else if not, then leave it at d (the old value). This is borrowed from C I believe. The cool thing in hansl (gretl scripting) is that it also works directly with matrices if you use the dot-operators. Example (which I cannot test right now): <hansl> matrix M = ( I(3) .> 0 ) ? -2 : 0 # note the .> print M </hansl> cheers sven _______________________________________________ Gretl-users mailing list -- gretl-users(a)gretlml.univpm.it To unsubscribe send an email to gretl-users-leave(a)gretlml.univpm.it Website: https://gretlml.univpm.it/postorius/lists/gretl-users.gretlml.univpm.it/